How Collaborative is your Robot? A Practical Approach
By Alberto Moel (Vice President Strategy and Partnerships)
When considering a human-robot collaborative workcell, we explicitly incorporate and expect humans to be safely working close to and/or interacting with the robot during operation. How this interaction takes place has been extensively analyzed 1 , and the potential approaches are as varied as the range of applications. Furthermore, the economics of human-robot collaboration can be compelling. 2
The conventional wisdom in human-robot collaboration is to use a Power and Force Limited (PFL) robot. However, PFL robots are, ahem, limited in payload and speed in order to remain safe upon human contact. Hence, if your application involves larger payloads, higher operating speeds 3 , or there are other hazards in the workcell or on the robot, the benefits of PFL are eliminated, and you might as well use a traditional robot.4 As we’ve pointed out previously, any robot out there can be made collaborative through the wonders of Speed and Separation Monitoring (SSM). 5
So, if you are using SSM, is there a metric of how collaborative is your application? Well, it depends on how collaborative your robot is. A point we have made before was that the closer the human could come to the robot before it needed to stop, the more collaborative we could make the application (and guarding could help with that). This has an intuitive logic: having the robot wait for a closer human approach before safely slowing makes the robot more “responsive” to human interaction. It also has economic benefits, such as lower cycle time (as the robot runs faster for a longer part of the cycle) and reduced workcell area.
This post will dive into the nitty-gritty on how we estimate this minimum safe human-robot distance, but if you’re pressed for time, the punchline is this: Based on manufacturer-provided stopping time and distance data (and under some very conservative assumptions), as long as the human is roughly 1.5 meters away from the robot, the robot can operate unimpeded at full speed. Conversely, the closest the human can approach before the robot is required to slow down or stop is about 1m.
These numbers, of course, vary by robot payload, reach, and speed (see Figure 2 and Figure 3 for a collection of our data), but in general, larger robots are as collaborative as small ones under this metric. And, practically, for a robot with a 2+ meter reach and moving at high speed, having to be a meter away from it is not an unreasonable request!
If you’d like to know for your specific robot what these distances are (and hence how collaborative it is) give us a shout here at Veo (sales@veobot.com or am@veobot.com) and we’ll gladly run the numbers for you.
Protective Separation Distance and Minimum Distance to Hazard
Now, we go into the weeds, if you’re still with us, dear reader. In our previous post we discussed at length the Protective Separation Distance (PSD), which is defined by ISO 10218-1 as
where:
Sp(t0) is the PSD at time t0;
t0 is the present or current time;
Sh is the contribution to the protective separation distance attributable to the human’s change in location;
Sr is the contribution to the protective separation distance attributable to the robot system’s reaction time;
Ss is the contribution to the protective separation distance due to the robot system’s stopping distance;
C is the intrusion distance, as defined in ISO 13855; this is the distance that a part of the body can intrude into the sensing field before it is detected;
Zd is the position uncertainty of the operator in the collaborative workspace, as measured by the presence sensing device resulting from the sensing system measurement tolerance;
Zr is the position uncertainty of the robot system, resulting from the accuracy of the robot position measurement system.
But there is another more “old fashioned” way to look at this distance, and for that we go way back to ISO 13855:2010 which deals with calculating safe distances between humans and running machinery. The scope of ISO 13855:2010 is “machinery,” and therefore this calculation considers a static hazard that can actuate (consider a saw blade or hydraulic press), but cannot change its position with respect to that of an operator in its vicinity. According to the standard, “the minimum distance to the hazard zone shall be calculated by using the general equation:”
where:
S is the minimum distance, in mm;
K is a parameter, in mm/sec, derived from data on approach speeds of the body or parts of the body;
T is the overall system stopping performance in seconds;
C is the intrusion distance, in mm.
Let’s call this distance S the Minimum Distance to Hazard, or MDH. If we generalize the above equation to make the robot the “hazardous machinery”, we see a correspondence between the MDH and PSD equations. The MDH calculation consists of two terms, which correspond to terms in the equation for the PSD calculation when safeguarding with SSM.
First the K x T term in the MDH calculation corresponds directly to S h, the contribution attributable to the human’s change in location, and the intrusion distance C corresponds identically to its counterpart in the PSD equation. The S s and S r terms are effectively zero in the MDH calculation because it assumes that the hazard itself does not move. The error terms Z d and Z r from the PSD calculation are not called out explicitly in the MDH calculation, but since they are constants, we can consider them subsumed as part of the intrusion distance C.
But the robot is moving, so how do we “map” it to the stationary machinery of ISO 13855? The way to do this is to introduce a Robot Future Cloud (RFC), which is all the space the robot could occupy by the time we have detected a PSD violation. We calculate the RFC to include a window of time before a PSD violation where the robot will continue to move through its planned trajectory, which may involve acceleration or deceleration.
The range of motion attributable to this pre-stop time window is calculated using the robot’s pose, the robot’s current load, the robot’s current velocity, a description of the robot dynamics and kinematics, any controller-enforced limits on robot joint velocity, and a bound on the time it takes for the sensing system (e.g. Veo FreeMove®) to issue a PSD violation signal and the time it takes the robot controller to acknowledge such a signal. The result of all this is a range of robot joint values that are attainable before the robot comes to a stop.
The range of joint values needs to be converted to a 3D space. This is done by comprehensively sampling the space of possible joint values and marking the space occupied by the robot’s links at these joint value samples. The result is an RFC representing all the 3D physical space the robot could possibly occupy before the sensing system assumes it could stop it. In terms of the PSD equation, the RFC is capturing the two terms missing in the MDH equation, S s and S r, and “incorporates” into the MDH calculation the fact that the robot is, in fact, moving.
Figure 1 shows visually what we are talking about. The yellow cloud captures the RFC, while the red cloud is the human approaching the moving hazard (the robot plus RFC).
In order to calculate the MDH, we need estimates for the RFC plus all the other terms in the PSD. We shall spare you the magic and will state without justification or evidence that C + Zd + Zr is 218mm (21.8cm). In other words, the PSD (and the MDH) is at least 21.8cm.
We also need estimates of the robot stopping times and distances. ISO 10218-1 Annex B requires that the stopping time and distance for different dynamic states of the robot (at a minimum, 33%, 66%, and 100% of speed, load, and extension, respectively) be provided by robot manufacturers. The requirement itself is interpreted differently by different robot manufacturers—some provide these data in tables of different sparseness, while others provide it in graphs that vary in resolution quality.
We also have the strong suspicion that the reported data are worst-case assumptions and extremely conservative. For example, ISO 10218-1 only states that data should be provided for those three speeds, the lowest being 33%, and there is no guidance around how to extrapolate down to zero speed. In some cases, the published stopping data indicates that it will take the same amount of time for the robot to stop whether it is moving at 33% speed with 33% load, 100% speed with 100% load, or any combination in between. This is likely because the worst-case number for 100% speed and 100% load was propagated throughout the table, and not because of robot dynamics that are insensitive to speeds or payloads.
Barring more complete and accurate estimates of these stopping times and distances, that is what we use for our calculations. To further handicap the calculations7, we make the worst case assumption that any occupied or potentially-occupied space contains a human and that a human in that space will move on a shortest-path toward the nearest space on the RFC.
How collaborative are most robots?
Since the time it takes the robot to stop is a factor in the human contribution to the PSD, robot stopping time and stopping distance are both critical. If the robot can stop quite close to its location at the moment the PSD is violated, but it takes a long time to come to a complete stop, the PSD will still be large because a human can move a large distance in that time.
At any rate, we would expect stopping times and distances to increase as payload, speed, and extension increase. Not all robot makers comply with providing extension data, and only provide stopping times and distances for 33%/66%/100% payload and speed at 100% extension. In other words, for each robot in our sample, we have a 3x3 matrix of stopping times and distances, with one axis 33%/66%/100% speed and the other one 33%/66%/100% payload, all at 100% extension.
Figure 2 shows our MDH calculations for a number of robots arranged by maximum payload, and Figure 3 shows the same arranged by maximum robot reach. For each robot, the lower value in the “error bar” corresponds to the MDH at 33% speed/33% payload and 100% extension, while the upper value is the MDH at 100% speed/100% payload and 100% extension. In other words, if the robot is moving slowly and lightly loaded, a human can be as close as the lower bar value before the robot has to begin slowing down or stopping. Conversely, a fully extended, fully loaded robot moving at full speed can continue moving at high speed as long as a human is further away from the upper value of the error bar.
A number of interesting details stand out from analyzing these charts:
The average MDH along all the robots, regardless of payload or reach, is about 1.3 meters, and is relatively tightly clustered around this value except for a couple of very large robots.
Similarly, the minimum MDH averages about 1 meter, and only a few robots have minimum MDHs much below this.
The range between the minimum MDH and the maximum MDH is also relatively narrow, about 50cm, although the range widens somewhat for the largest robots.
Some robots have surprisingly short MDHs, even for large payloads, which may indicate they are better mechanically suited for collaborative applications.
Some robots have a narrow (sometimes even zero) range between the upper and higher MDH bounds. This would result from the 3x3 stopping time and distance tables having the same values across the board. Whether this is a result of conservatism (propagating worst case values across the table) or robot geometries (making them relatively insensitive to speed and payload variations) is a subject of further analysis.
What do we take away from our quantitative dive into SSM collaborative robotics? First, it is that most robots can be pretty collaborative and being large is not a major disadvantage. Second, relative to the size of some of these robots (a 2200mm reach requires a minimum 4.4m footprint), a meter or a bit more for the human to stand safely is not that much, really. Lastly, these are some very conservative estimates. Using more realistic application-based values (who runs at 100% extension?) means it is very likely humans can stand very close to robots moving rapidly and still remain safe under the stringent standards.
1See for example, Bdiwi et al, A new strategy for ensuring human safety during various levels of interaction with industrial robots or Federico Vicentini’s Terminology in safety of collaborative robotics for a good review of practical human-robot collaboration.
2We have looked into this question before, where we found benefits in lower capital expenditures, faster cycle times, shorter fault recovery, and lower reconfiguration costs.
3Many times, a necessity to increase cycle times in order to make the economics of the application “work.”
4Ranging from the workpiece, the end effector, or other dangerous equipment in the workcell.
5This fact is not adequately appreciated. All forms of ISO 10218-1 certification for collaborative workcells require that the robot itself have 3rd-party certified Category 0 and Category 1 stops per ISO 13850 or ISO 60204-1. Every robot sold commercially has 3rd-party safety certified Category 0 and Category 1 stops, which are key building blocks in SSM collaborative applications and systems, including Veo FreeMove®.
6 But trust us, we’ve given this more than necessary thought. And the logic is available upon request, if you are that nosy.
7Stay tuned on that, it is a major initiative at Veo Robotics to come up with better and more precise estimates of these stopping times and distances.